Awesome Ai

Generative AI can be used for good or evil. In the case of Adobe Premiere Pro, we’re excited about the potential of its AI tools to improve video editing workflows with the ability to add or remove objects with just a couple of clicks and extend scenes if they don’t fit your timeline. They also are testing integration with models like Sora AI to generate entire scenes.

Advancements in voice recognition, large language models, neural networks, and robotics come together in this impressive demonstration. An unnamed human asks the Figure 01 robot to recognize objects, perform tasks, and explain its reasoning. While it takes a moment for the AI to compute its responses, it’s a glimpse of things to come. (Thanks, Myriam!)

Honest Ads offers a darkly humorous take on what could happen if our worst fears about artificial intelligence come true. Along the way, fake CEO Roger Horton exposes the kind of machine-generated garbage that passes for “content” these days and how large language models are just really good at guessing what the next word might be in a sentence.

Artificial intelligence has become a major subject of today’s technology conversations. In 1983, one company marketed an early smart home product designed to control home appliances via voice control like Siri or Alexa. Kevin from Popular Science got his hands on the costly Butler in a Box to explain how it worked without Wi-Fi, Bluetooth, or the Internet.

AI voice generation software has made it relatively easy to create songs and speech that sound like anyone. But can an AI singer beat a professional like Anthony Vincent? The Ten Second Songs host asks you to be the judge. While the technology certainly is impressive, we will always be Team Human when it comes to creative pursuits.

AI tools like ChatGPT are incredibly powerful. Backed by large language models, these systems can create convincingly detailed text, write code, and answer questions. Interaction designer Arvind Sanjeev turned a Brother electric typewriter into an interface to GPT-3. Users can type a prompt, and the AI types its response on the same sheet of paper.

This impressive yet creepy AI tech can take a single still image and make it appear to speak or sing any audio. Developed by researchers from Alibaba’s Institute for Intelligent Computing, EMO: Emote Portrait Alive’s model was trained on more than 150 million images to work its magic. The video is peppered with examples of the convincing deepfakery.

AI Warehouse has used machine learning to train figures to walk and run. This time, they’ve given Albert the bot a much more significant challenge – climbing stairs. Before the blockhead can work on that, he must learn to plant his feet on a step without falling down. After the staircase, things get more complex with uneven terrain and an escalator.

Karma Police, arrest this girl; her Britney hairdo is making me feel ill. There I Ruined It used AI voice matching tech to make Radiohead’s Thom Yorke sing the lyrics to Britney Spears’ Baby One More Time and Oops, I Did It Again to the tune of Karma Police. It sounds like a terrible idea, but it’s way better than we thought it would turn out.

Tech startup Rabbit turned up at CES 2024 with a pocket-size virtual assistant. Designed in collaboration with Teenage Engineering, the gadget can rapidly answer questions using AI. It has a push-to-talk button, a 2.88″ touchscreen, an analog scroll wheel, two microphones, a speaker, and a rotating camera. The demo kicks off at 6:33 in the video.

The use of artificial intelligence to generate and manipulate images has had its share of controversy, but recent technological advancements are still amazing. Two Minute Papers explores four very different ways to use AI: making still images dance, compressing video conferencing, upscaling video resolution, and making software that can identify smells.

Builder John Malecki wanted a new media center for his office. Rather than designing the cabinetry himself, he turned to generative AI to come up with the concept. After playing with a few prompts, Midjourney came up with a design. Its organic shapes caused construction challenges, but John and his team did a great job bringing the real-world version to life.

We have robotic musicians, cooks, and barbers. Engineers at the University of Tokyo taught a robot to play Charades. The team took the Alter3 humanoid robot, connected it to OpenAI’s GPT-4, and used what it learned to act out prompts through its movements. Check out more of the bot’s interpretive dances on Takahide Yoshida’s YouTube channel.

ChatGPT is one of the most talked-about subjects of the decade. Despite its questionable accuracy, it does an amazing job of creating human-sounding text. Squirrel Monkey set their time machine to 1988 to see what the AI might have been like had OpenAI released the tech 40 years prematurely. Offline mode would have been particularly awful.

What happens when you ask an artificial intelligence to write a script for a short film about AI’s impact on society? You get this dystopian nightmare. Nobody & The Computer posted this AI-generated movie, which imagines what the future might look like once we all have AI-powered augmentations and robots have taken all of our jobs. (Thanks, Rob.)

The term “chatbot” is typically associated with those frustrating customer service AI gatekeepers before chatting with a real human. But in the case of this Boston Dynamics’ Spot robot, the chatbot has four legs and can walk. Engineers from the robotics company used ChatGPT and other AI models to enable Spot to answer questions and act as a tour guide.

Proving you must crawl before you walk or run, AI Warehouse tasked a group of five AI agents with completing a 100-meter dash. Each was trained using Deep Reinforcement Learning to reward them for correct behaviors and punish them for incorrect ones. Eventually, they get their boxy butts off the line and wobble their way through the course.

After a failed attempt to do a backflip in real life, animator Nikita Diakur thought it would be safer to have a digital stand-in perform the maneuver. With the help of engineer Maximilian Schneider and machine learning tech based on the paper Deep Mimic, he attempted to nudge his avatar in the right direction to achieve his gymnastic goal.

And the court will rise, while the pillars all fall. Peter Gabriel has been creating musical masterpieces since the 1960s. The Court comes from his new album i/o. The track explores society’s need for order and the flaws of judicial systems. The video was created by Junie Lau using various AI tech, including Stable Diffusion, ChatGPT, MidJourney, and DALL·E 2.

(PG-13: Language) Every band has its fans and detractors. There I Ruined It created this song for people who don’t like the Red Hot Chili Peppers. The track was made using AI tech to emulate Anthony Kiedis’ and John Frusciante’s vocals – but this version of Snow (Hey Oh) has some new (and improved?) lyrics.

FRVR’s AI game development platform will let users create their own video games using text prompts. The demo shows it taking a description of a top-down shooter and generating the code for a playable game. As the user refines their prompts, the game gets better. Users will be able to publish and monetize the games they create too. Full walkthrough video here.

(PG-13: Language) As artificial intelligence tech gets more capable, there are many ethical questions that humanity will have to deal with. Among them is how we will treat AI if and when they become conscious. Philosopher exurb1a explores the challenges we’ll face figuring that out, along with some of the many dangers of AI tech.

Voice mimicry is getting easier and easier to achieve thanks to rapid advancements in AI tech. Musician Kid Klava shows just how impressive the technology has become by taking a Lennon-inspired song he wrote and programming the late Beatles’ digital avatar to perform it.